A multi-task Gaussian processes method for risk identification during cable-stayed bridge cable replacement

• Developed a dynamic safety evaluation and forewarning method for bridge cable replacement. • Developed real-time methods for cable force and girder deflection identification during cable removal. • Developed MTGP method for data reconstructing and predicting to address the inevitable data missing issue. • Proposed three risk probability indices for structural safety evaluation during cable replacement. • Validated the proposed method using a real cable replacement project of a 24-year-old cable-stayed bridge. Cable replacement is essential for extending the service life of cable-stayed bridges but poses significant safety risks due to the concurrent deterioration of other structural components. Ensuring safety during rapid cable detensioning requires timely assessment of the evolving structural states. This study proposes a multi-task Gaussian process (MTGP)-based dynamic evaluation and risk identification method for bridge safety during cable replacing process. A quasi-real-time cable force identification method and a vision-based deflection monitoring approach are first adopted to capture high-resolution structural responses. To mitigate inevitable data gaps, the MTGP dynamically reconstructs missing responses at each unloading phase, ensures spatial completeness, and further predicts structural behaviour under forthcoming adverse unloading scenarios. Based on the MTGP’s probabilistic outputs, risk assessment indices are formulated for both cables and girders, enabling quantitative evaluation of component safety and issuing early warnings. Validation on a practical cable replacement project demonstrates that the MTGP achieves accurate data reconstruction and reliable prediction of structural responses. The proposed indices effectively identify safety conditions and quantify risks in subsequent unloading phases. Overall, the method advances dynamic, learning-based safety evaluation for cable replacement operations, offering a practical framework to support decision-making and risk management in bridge maintenance.